Hi all,
I have recently encountered an issue with PTS 1.6.0 in the parallel bzip2 benchmark. It uses all available RAM (2Gb + 3Gb swap). I caught it before anything bad happened and killall'd the pbzip2 process. My system was crawling by this stage. My swap partition is on a 74Gb Raptor. PTS is installed on a 250Gb Seagate.
Below is my system info as reported by PTS:
phoronix-test-suite system-info
====================================
Phoronix Test Suite v1.6.0 (TYDAL)
System Information
====================================
Hardware:
Processor: AMD Athlon 64 X2 Dual Core 6000+ @ 3.00GHz (Total Cores: 2), Motherboard: MSI MS-7250, Chipset: nVidia MCP55, System Memory: 2008MB, Disk: 250GB ST3250624NS + 80GB WDC WD740ADFD-00NLR1 + 250GB ST3250823AS, Graphics: GeForce 7900 GT/GTO 256MB (450/660MHz), Screen Resolution: 1680x1050
Software:
OS: Ubuntu 8.10, Kernel: 2.6.27-12-generic (x86_64), Desktop: GNOME 2.24.1, Display Server: X.Org Server 1.5.2, OpenGL: 2.1.2 NVIDIA 177.82, Compiler: GCC 4.3.2, File-System: ext3
Now, suggestions. We've been talking about PTS and benchmark results over at the AMDZone forums, and some posters have made some suggestions.
Firstly, I made the suggestion of a standardised Phoronix Live CD. It was seconded by another poster at AMDZone. I know this has been mentioned in various ToDo lists here. I believe it's very important for PTS to be more robust in its reproducibility of results. Having a standardised live CD environment (obviously being able to specify where PTS should be installed on the host's filesystems) would be a huge step up for PTS.
Another poster at AMDZone suggested par2 to be part of the benchmark suite. A single threaded test, but still interesting none the less.
Lastly, on PTS Global, I would find it very useful to be able to search for individual CPU models plus benchmarks at the same time. For example, find all GnuPG benchmarks on Core 2 Quads and Phenoms (heh, gotta love the strength of AMD CPUs in encryption ). This would make it much easier to compare different CPUs. Sure it's fairly apples-to-oranges with peoples' systems, but it gives a good idea. Being able to search simultaneously for system property and benchmark type would make the process quicker.
That's all for now. Thanks Michael & team for the great work on PTS. It would be great to see mainstream review sites (other than Phoronix) using PTS regularly to test new hardware. I believe it will get there, or at least PTS will be perfectly suitable for such use (as a standardised Live CD).
I have recently encountered an issue with PTS 1.6.0 in the parallel bzip2 benchmark. It uses all available RAM (2Gb + 3Gb swap). I caught it before anything bad happened and killall'd the pbzip2 process. My system was crawling by this stage. My swap partition is on a 74Gb Raptor. PTS is installed on a 250Gb Seagate.
Below is my system info as reported by PTS:
phoronix-test-suite system-info
====================================
Phoronix Test Suite v1.6.0 (TYDAL)
System Information
====================================
Hardware:
Processor: AMD Athlon 64 X2 Dual Core 6000+ @ 3.00GHz (Total Cores: 2), Motherboard: MSI MS-7250, Chipset: nVidia MCP55, System Memory: 2008MB, Disk: 250GB ST3250624NS + 80GB WDC WD740ADFD-00NLR1 + 250GB ST3250823AS, Graphics: GeForce 7900 GT/GTO 256MB (450/660MHz), Screen Resolution: 1680x1050
Software:
OS: Ubuntu 8.10, Kernel: 2.6.27-12-generic (x86_64), Desktop: GNOME 2.24.1, Display Server: X.Org Server 1.5.2, OpenGL: 2.1.2 NVIDIA 177.82, Compiler: GCC 4.3.2, File-System: ext3
Now, suggestions. We've been talking about PTS and benchmark results over at the AMDZone forums, and some posters have made some suggestions.
Firstly, I made the suggestion of a standardised Phoronix Live CD. It was seconded by another poster at AMDZone. I know this has been mentioned in various ToDo lists here. I believe it's very important for PTS to be more robust in its reproducibility of results. Having a standardised live CD environment (obviously being able to specify where PTS should be installed on the host's filesystems) would be a huge step up for PTS.
Another poster at AMDZone suggested par2 to be part of the benchmark suite. A single threaded test, but still interesting none the less.
Lastly, on PTS Global, I would find it very useful to be able to search for individual CPU models plus benchmarks at the same time. For example, find all GnuPG benchmarks on Core 2 Quads and Phenoms (heh, gotta love the strength of AMD CPUs in encryption ). This would make it much easier to compare different CPUs. Sure it's fairly apples-to-oranges with peoples' systems, but it gives a good idea. Being able to search simultaneously for system property and benchmark type would make the process quicker.
That's all for now. Thanks Michael & team for the great work on PTS. It would be great to see mainstream review sites (other than Phoronix) using PTS regularly to test new hardware. I believe it will get there, or at least PTS will be perfectly suitable for such use (as a standardised Live CD).
Comment